Objectives Patients who are deteriorating, with uncertain recovery and with a short prognosis often have complex needs. The AMBER care bundle systematically manages these patients by promoting consistent communication and care planning. To describe how the AMBER bundle is applied in a UK hospital. To identify factors affecting the use of the AMBER bundle in patients who subsequently died. To gain preliminary data with regard to potential markers of deterioration in groups less frequently supported by AMBER

Methods Retrospective review of electronic case notes for all deaths over 11 months on five inpatient wards where AMBER was implemented.

Results N=149. Median age 80, IQR 72–87. Admission diagnoses: cancer (25%), non-cancer (31%), multimorbidity (44%). 38% were supported by AMBER. Patients with cancer were more frequently supported by AMBER (62% vs 30% p<0.001). Illness trajectory was defined a priori as ‘predictable gradual’ (40%), ‘predictable rapid’ (22%), ‘unpredictable’ (21%) or ‘sudden death’ (17%). ‘Predictable gradual’ deterioration resulted in more frequent support by AMBER (62% vs 21%, p<0.001). Patients were supported by AMBER after median 10 days of admission (IQR 5–17 days), and died median 9 days later (IQR 3–15 days). Patients with multimorbidity or unpredictable deterioration were less frequently supported by AMBER. Potential markers of deterioration were acute kidney injury, delirium, falls and comorbidity.

Conclusions The use of the AMBER care bundle is affected by illness trajectory and diagnosis. Future work should clarify predictors of deterioration in patients with an unpredictable course.